Traffic Signal Coordination and Queue Management in Oversaturated Intersection

摘要

Traffic signal timing optimization when done properly, could significantly improve network performance by reducing delay, increasing network throughput, reducing number of stops, or increasing average speed in the network. The optimization can become complex due to large solution space caused by many combinations of different parameters that affect traffic operation. In this study three different methods are used to find near-optimal signal timing parameters in transportation networks. The methods are: Genetic Algorithms (GA), Evolution Strategies (ES), and Approximate Dynamic Programming (ADP). Each method is introduced, the signal timings associated with them are explained and some important measures of performance of the networks are determined and compared. One small network with 9 intersections and one medium network with 20 intersections were used for evaluating the optimizations methods. Three general cases are discussed in this report. For the small symmetric network, three levels of traffic loading are used (no overload, 10% overload and 20% overload). For the medium network (modified Springfield IL downtown network), two levels of entry volumes are used (750 and 1000 vehicle per hour per lane).